n8n v2.0 AI Workflow Platform

Overview

n8n v2.0 is the first stable general availability (GA) release of the leading open-source workflow automation platform, officially launched on July 24, 2026. With over 58,000 GitHub stars and a massive ecosystem of 1000+ integrations, n8n v2.0 represents a milestone in making AI workflow automation accessible to teams of all sizes.

The v2.0 release introduces the AI Workflow Canvas, a native visual interface for building AI workflows with LLM nodes (OpenAI, Anthropic, local Ollama), vector retrieval nodes, and pre-built RAG pipeline templates. This transforms n8n from a traditional automation tool into a comprehensive AI workflow platform.

Features

AI Workflow Canvas: Native LLM nodes with OpenAI, Anthropic, and local Ollama support
Vector Retrieval Nodes: Built-in vector database nodes for semantic search
RAG Pipeline Templates: Pre-built retrieval-augmented generation workflows
Data血缘图 (Data Lineage): Visualize data flow and field mappings across nodes
Cost Estimator: Per-node token/API cost display for AI workflows
1000+ Integrations: Connect with virtually any external service
Visual Workflow Builder: Drag-and-drop interface with error handling
Enterprise Features: Version control, team collaboration, access management

Installation

# Install via npm (recommended)
npm install -g n8n

# Start n8n
n8n start

# Or via Docker
docker run -d --name n8n -p 5678:5678 -v n8n_data:/home/node/.n8n n8n/n8n

# For local model support (Ollama)
docker run -d --name ollama -p 11434:11434 ollama/ollama

Quick Start

Create Your First AI Workflow

Launch n8n: n8n start and open http://localhost:5678
Create New Workflow: Click "Create New Workflow"
Add Trigger Node: Select "Schedule Trigger" or "Webhook Trigger"
Add LLM Node: Search for "OpenAI" or "Anthropic" and add to canvas

Configure LLM Node:

Model: gpt-4
Prompt: Summarize this article: {{ $json.article_text }}
Temperature: 0.7

Add Output Node: Send results to Slack, email, or API

Example: AI News Summarizer Workflow

# Workflow: AI News Summarizer
trigger: "Webhook"
steps:
  - node: "HTTP Request"
    action: "GET"
    url: "https://newsapi.org/v2/top-headlines"
  - node: "OpenAI"
    action: "Summarize"
    prompt: "Summarize the following news articles in 3 bullet points"
    input: "{{ $json.articles }}"
  - node: "Slack"
    action: "Send Message"
    message: "{{ $json.summary }}"

Core Concepts

AI Workflow Canvas

The AI Workflow Canvas is a specialized visual editor for AI workflows with:

Drag-and-drop LLM nodes: OpenAI, Anthropic, Ollama, and custom model nodes
Vector database nodes: Connect to Pinecone, Qdrant, Chroma, or local vector stores
RAG templates: Pre-built workflows for document retrieval and generation
Agent nodes: Deploy full AI agents as workflow steps

Data Lineage Visualization

The new "Data Lineage Graph" shows:

Input → Transform → LLM → Filter → Output
  ↓        ↓        ↓      ↓       ↓
Field:    Field:   Token:  Count   Payload:
text      text     512     42      JSON

Cost Estimator

The enterprise cost estimator shows per-node costs:

Node	Input Tokens	Output Tokens	Cost (USD)
OpenAI GPT-4	1,000	500	$0.03
Anthropic Claude	800	400	$0.02
Vector Search	N/A	N/A	$0.001
Total	1,800	900	$0.051

Advanced Features

RAG Pipeline Template

# Pre-built RAG pipeline
workflow: "Document Q&A RAG"
nodes:
  - name: "Document Loader"
    type: "File Upload"
  - name: "Text Splitter"
    type: "Document Splitter"
    chunk_size: 1000
    chunk_overlap: 200
  - name: "Embedding Generator"
    type: "OpenAI Embeddings"
    model: "text-embedding-3-small"
  - name: "Vector Store"
    type: "Pinecone"
    index: "documents"
  - name: "Retriever"
    type: "Vector Search"
    top_k: 5
  - name: "LLM Generator"
    type: "OpenAI"
    prompt: |
      Based on the following context, answer the question:
      Context: {{ $json.retrieved_documents }}
      Question: {{ $json.question }}
  - name: "Response Formatter"
    type: "Template"
    output: "{{ $json.answer }}"

Agent Node Configuration

node: "Agent"
config:
  type: "ReAct"
  model: "gpt-4"
  tools:
    - "search_web"
    - "read_file"
    - "write_file"
  max_iterations: 10
  prompt: "Research the topic and write a report"

Error Handling in AI Workflows

workflow: "AI Content Generator"
error_handling:
  retry_policy:
    max_retries: 3
    backoff: "exponential"
    retry_on:
      - "rate_limit_exceeded"
      - "temporarily_unavailable"
  fallback:
    on_failure: "Use cached response"
    cache_ttl: "24h"

Dynamic Workflow Execution

# Dynamic node execution based on AI output
def execute_workflow(context):
    ai_output = call_llm_node(context.get("input"))
    
    if ai_output.get("classification") == "urgent":
        return execute_urgent_workflow(ai_output)
    elif ai_output.get("classification") == "normal":
        return execute_normal_workflow(ai_output)
    else:
        return execute_batch_workflow(ai_output)

Local AI Integration (Ollama)

# Local model workflow
nodes:
  - node: "Ollama"
    model: "llama3.1:8b"
    prompt: "Analyze this data: {{ $json.data }}"
  - node: "Ollama Embeddings"
    model: "nomic-embed-text"
    text: "{{ $json.document }}"

Examples

Example 1: AI-Powered Data Pipeline

# Python workflow with n8n API
import requests

workflow_id = "123"
payload = {
    "data": {
        "input": {
            "data": {"articles": [...]}
        }
    }
}

response = requests.post(
    f"http://localhost:5678/workflows/{workflow_id}/execute",
    json=payload
)

Example 2: Multi-Step AI Content Pipeline

Research: LLM node gathers information from web search
Outline: LLM node creates structured outline
Write: LLM node generates content section by section
Review: LLM node reviews and improves content
Publish: API node publishes to WordPress/GitHub

Example 3: AI Customer Support Workflow

Trigger: Webhook receives customer query
RAG: Retrieve relevant knowledge base articles
Classification: LLM classifies query type
Routing: Route to appropriate agent or knowledge base
Response: Generate personalized response

Pros

✅ 1000+ Integrations: Largest integration library in workflow automation
✅ AI-Native Canvas: Purpose-built for AI workflow development
✅ Visual Debugging: Data lineage graph for easy troubleshooting
✅ Cost Estimator: Enterprise token cost visibility
✅ Self-Hostable: Full data control with Docker deployment
✅ Active Community: 58,000+ GitHub stars, large ecosystem

Cons

✅ Complex Learning Curve: Many features can overwhelm beginners
✅ Resource Intensive: Requires significant memory for large workflows
✅ Fair-Code License: Commercial use requires paid license
✅ AI Feature Maturity: AI nodes added in v2.0, less mature than core automation

When to Use

Enterprise AI Workflows: Complex AI pipelines with multiple integrations
RAG Applications: Document retrieval and generation workflows
Multi-Step Automation: Workflows combining AI and traditional automation
Cost-Conscious AI: Token cost visibility and optimization
Self-Hosted AI: Local model support with Ollama
Team Collaboration: Multi-user workflow development and deployment

n8n v2.0 AI Workflow Canvas

Workflow Steps

Download

Documentation